不确定状况下的判断（烂尾）

原书及其作者

这本《不确定状况下的决策：直觉与偏见》，是我从 Ignatius Lee 李聿脩 的一篇note里看到的。两位作者在研究人的决策机制的问题上，大大挑战了此前占据统治地位的理性人模型，提出了一个 直觉／反思的双元模型（看不明白不要紧，这系列的笔记会逐步展开，这个概念目前没有太通用的名字）。此举直接开辟了一个新的研究领域，行为经济学，也为两位作者赢得了诺奖。

但是很遗憾，这本书并不是 Amos Tversky 和 Daniel Kahneman 原本的重量级之作，这本书是对他们研究结果的分析，写的非常之无聊，我实际上是投错门了。能让人感到安慰的是，这本书很短，只有不到一百页。而且虽然文字无聊，每章给出的 key points 基本覆盖一章的所有文字内容，只看这部分提要的话可以获得对两位原作者的学术成果的快速概览和背景了解。

---

An Analysis of Amos Tversky and Daniel Kahneman’s Judgment under Uncertainty: Heuristics and Biases

Section 1: Influences

P27 The debate about rationality and choice was at its most fierce in the field of economics. While rational agent theorists did not claim that the theory can give a perfect account of how we make decisions, they did argue that the model was useful for predicting and comparing the outcome and pattern of choices. So the question became: “How far can we push the rational agent model?”

p27 In 1953 the French psychologist Maurice Allais warned that the mathematical formalism of economics literature hid many key psychological constraints that affect decision-making, emphasizing the role of psychology: “What an individual considers in a random choice is not the monetary value of a gain but the psychological value attached to that gain.”

• 在写这篇论文的两个作者之前，经济学界对于人的行为预测是基于理性人假设（rational agent），在此假设框架下发展起来的经济学流派叫 Homo economicus。这个假设认为个体有无限的正确认知能力、充分的资讯、会做出最大化自己利益的决策，此为 “理性”。尽管经济学家也知道不现实，但这一直以来还是被视为一个能够方便的对个体行为给出大致预测的好工具。
• 原论文的两位作者，Amos Tversky 和 Daniel Kahneman，主要关注人在后果不确定、资讯不完全的情况下的决策机制——即标题 “Judgement under Uncertainty”。可以注意到这从一开始就出了理性人假设的前提了，但这篇论文在问世之后的主要影响，正是让人们开始意识到，理性人假设和现实世界到底偏了有多远、以往是否被信任过头了。而其原因，就是两位作者的关键发现：人在不确定状况下的决策，对理性人模型存在着系统性的偏差。
• 系统性偏差是和偶然误差相对的一个概念，偶然误差是不可避免的，但是一般而言是随机的；系统误差指的是由于实验设计有缺陷等等，导致结果存在一个普遍的偏差，比如说预测结果永远比实际结果要大，那可能就预示了某种系统性偏差的存在。

p30 Tverksy and Kahneman took a novel approach to their work, using experimental questions to study decision-making. That is, instead of tackling the problem from a theoretical angle, they created experimental questions and puzzles for people to solve. …… (两位作者表示) My model for this kind of psychology was research reported by [the US psychologist] Walter Mischel in which he devised two questions that he posed to samples of children in Caribbean islands: ‘You can have this (small) lollipop today, or this (large) lollipop tomorrow,’ and ‘Now let’s pretend that there is a magic man … who could change you into anything that you would want to be, what you would want to be?’ The responses to these lovely questions turned out to be plausibly correlated with numerous characteristics of the child and the child’s background. I found this inspiring: Mischel had succeeded in creating a link between an important psychological concept and a simple operation to measure it.

p36-37 While Tversky and Kahneman’s “Judgment under Uncertainty” appeared in a scientific journal, it is not filled with highly technical jargon. …… Tversky and Kahneman do a masterful job of providing specific, concrete examples of the points they are trying to make. These examples tend to engage readers and allow them to think about what their answer would have been. When the correct answer is described, they see for themselves how their own judgment is biased in systematic ways. Such personal demonstrations make Tversky and Kahneman’s points in a salient, simple, and powerful way.

• 关于修辞和实验方法的事，最好还是自己看原文，二手总结传递不了精华。

Section 2: Ideas

p34-35 Heuristics are cognitive tricks that our minds use—automatic, simple ways to make decisions in complex contexts. They are essentially mental shortcuts or rules of thumb. Although heuristics might fail unpredictably, sometimes they fail in ways that are predictable.These predictable failures of heuristics are known as biases; they often predispose us toward making incorrect decisions. …… most heuristics, its validity is limited to specific settings.

p38-39 These biases affect the day-to-day decisions of both laypeople and experts—having a possible negative impact on scientific research. …… when they think intuitively.

p40-41 The two researchers put together the “dual model” of thinking, according to which humans can make decisions in two ways—a fast, intuitive way, prone to systematic biases and errors, and another reflective, slower way that is less influenced by mental shortcuts and systematic errors.

• 这大概会是全书最重要的概念，也是标题里的直觉与误差（Heuristics and Biases）。人类每天都要做决策，但是让人的大脑去认真推理是一件很耗能的事情，所以大多数时候的人其实只是在模拟推理——好像思考了，但其实是迅速组装固有常识。两位作者对于其时的学界，非常重要的一点贡献就是提出了这个双元的决策模型：直觉式和反思式（intuitive vs reflective）。
• Biases 一般译为偏见，但我更想强调它作为一种道德中性的 “只是偏差” 的意思，故我在这系列笔记里将会无差别的使用三种译法：偏见、误差、偏差。

p35-36 The three heuristics presented in the paper are representativeness, availability, and adjustment and anchoring. …… To understand the heuristic of representativeness, consider that we judge things and people based on how representative of, or similar they are to, stereotypes. …… Availability concerns the way in which people judge how often an event occurs based on how readily they can remember a similar event. …… Adjustment and anchoring concerns numbers.We tend to overweight the first numbers we see; this is “anchoring.” Then, when we gain subsequent information, the adjustments we make tend to be insufficient.

• 两位作者把自己观察到的直觉分为三类，Representativeness, Availability, Adjustment and Anchoring。第一个好理解，就是当我们观察事物时，常会倾向于把这个事物套到刻板印象里去理解，从而基于典型的印象来对当下我们所观察的个体做出预测；第二个结合原文里的具体例子其实更好理解，说的是我们容易记住的东西就会觉得它好像更常发生，比方说每天的日常都在发生、但是没什么记忆点的事件的出现频率可能被低估，而更少发生、但能让人记忆深刻的事情的出现频率可能会被高估；最后一个结合例子就很好说了，比方说让人立刻估计 1x2x3x4x5x6x7x8 和 8x7x6x5x4x3x2x1 的结果的话，很多人会下意识低估前者而高估后者，这就是最先看到的数据在我们心里形成的锚定效果。

p38 Sometimes people make biased decisions because they are motivated to do so. For instance, according to the bias of “wishful thinking,” people form beliefs that please them, regardless of evidence to the contrary. But Tversky and Kahneman also demonstrate that we make errors of judgment even when we are motivated not to do so. They found that such errors were made even when participants were financially rewarded for giving the correct answer.

• 就我所认为的，导致人们在本该精准决策、但还是使用了直觉式的错误决策的原因大概就有这么几类：锁死，无意识，隐性激励。当我说 “锁死” 的时候，是指有些时候比方说一个人自己钻牛角尖，往一个方向上越想越不明白，但是又无法自己脱离这种状态，一遍一遍的重复错误决策；无意识就是想表示，真的没注意到自己的决策链路受到了偏见的影响，哪怕被纠正了也可能还是万分单纯的意识不到，所导致的重复出错；隐性的激励是指，虽然明面上看起来应该正确决策比较好，但是其实这个人通过重复做出错误的决策，获得了比方说自我优越感、或者证明自己之前的决策没有错、或者将自己塑造成了想要的社交形象之类的，在实验条件以外的激励。
• 当然，不能忘了最朴实无华的 “bounded rationality”，有限理性。这也是在这篇论文以前，另一位美国的政治学与社会学家 Herbert Simon 提出的概念，顾名思义就是说有时候人就是没有足够的认知能力去完成一些任务，比方说给出一则很复杂的运算，或者是一个非常抽象的数论证明题，非要普通人立刻给出答案，那除了瞎猜也没什么办法，重复再多次也只能瞎猜。

p39-40 People “regard the sequence H-T-H-T-T-H to be more likely than the sequence H-H-H-T-T-T, which does not appear random, and also more likely than the sequence H-H-H-H-T-H, which does not represent the fairness of the coin.” In other words, people seem to expect small sequences of events to reflect the overall pattern of those events. Furthermore, they tend to be perfectly willing to make predictions about some future outcome based on intuition and limited information—predicting, for example, that a company will be profitable in the future when given a positive description of that company, even if the description contains no information about profits.

• 我觉得，人总会遇到讯息不足但是必要做决策的情况，而且频率不会低。这种时候，如果没有一个 “有效性差不多就行了，重要的是要能在各种奇葩环境下运行的又快又好的” 决策机制的话，每次都需要做大量的调查、研究，然后 “理性” 决策的话，人的脑子会不够用的——时间、精力、认知资源、可获取的资讯范围无一不在限制着这种理想情况的出现。所以这也许就是人们为什么如此的适应于通过极小的样本量，在甚至是可能本不存在关系的地方构建出相关性推理。

Section 3: Impact

• Amos Tversky 和 Daniel Kahneman 的整个研究历程，起于他们合作发表的两篇关键论文，1974年发布于 Science 期刊的 Judgment under Uncertainty: Heuristics and Biases，一个是1979年发布于 Econometrica 期刊的 Prospect Theory: An Analysis of Decision under Risk。这两篇论文都不会太难读，尤其是第一篇只有8页，第二篇也只有30页。两位作者在当时也处于他们合作生涯的相对早期，研究并不深入，但是他们的理论已经体现出了无穷的潜力，一石激起千层浪。前者关于人的普遍决策机制，提出了挑战理性人模型的双元决策模型；后者基于这一工作继续发展，关于人在面对风险和不确定情况时的决策机制的问题，提出了挑战此前的功效理论（utility theory）的展望理论（prospect theory）。这两篇论文是一切最初的起点，包含两位作者最原初的研究工作。
• 接下来，Cambridge University Press 于1982年出版了一部同名专书，Judgment under Uncertainty: Heuristics and Biases两位作者不仅在更多年的合作后进一步完善了自己的理论，并且也收集整理了更多证据包含在这一本学术著作中。这本书大概五六百页。
• 后来，两位作者之一的 Amos Tversky 在1996年逝世，Daniel Kahneman 则在2011年再通过 Doubleday Canada 出版了对他们两人合作生涯的一部更加通俗的综合反思与总结，这就是畅销书 Thinking, Fast and Slow。

---

原书信息

• Kahneman, D. (2011). Thinking, fast and slow. New York Farrar, Straus & Giroux Inc.
• Kahneman, D., & Tversky, A. (1974). Judgment under Uncertainty : Heuristics and Biases (pp. 1124–1131). Cambridge University Press.
• Kahneman, D., & Tversky, A. (1979). Prospect theory: an Analysis of Decision Under Risk. Econometrica, 47(2), 263–292. https://doi.org/10.2307/1914185
• Morvan, C. (2017). Judgment under uncertainty - heuristics and biases. Macat International Limited.
• Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and Biases. Science, 185(4157), 1124–1131.